Wednesday, February 18, 2009

The Negative Convexity Of CDS Trading And Why CDOs Chase Markets

Anyone who has ever traded CDS has noticed the self-fulfilling prophecy of accelerated widening or tightening on an initial move in single name spread. This phenomenon has stumped traditional cash credit investors who don't realize the peculiar technicalities of the CDS market. It is explained by dealer structured credit desks that tend to chase the market, meaning as spreads widen they buy CDS, and vice versa. Because of this, otherwise small moves in single names can lead to profound jumps in spread.

How does it work:
  • A client sells a synthetic CDO tranche to a structured credit desk. The desk is now long risk in the CDO. From a transaction perspective, the client bought CDO CDS while the desk sold the protection.

  • The dealer has to offset risk and can either buy protection on the same CDO tranche or hedge by proxy thru single-name CDS. As the CDO world is limited and at times very illiquid, usually the easiest trade is to use single-names as hedges.

  • Dealers use models to estimate how much a CDO tranche should move for a given move in the underlying spread. The below chart shows amount of single-name CDS a desk needs to buy in order to hedge $100 mm long risk from CDO position. As spreads widen, the dealer has to buy more and more protection to hedge (i.e. negative convexity).

  • Assume the correlation desk bought CDO when spread was 150 bps. To hedge, the dealer bought $250 mm single-name CDS. Now the portfolio spread moves 50 bps wider so average single-name spread is 200 bps. Dealer looks at the model and realizes $250 mm hedge is too small and now needs $300 million, which difference ($50 million) in the single-name he promptly goes to the market and buyss, thereby pushing prevailing offers even wider in the process, starting a feedback loop for other CDO managers in the market. As spreads continue to widen, the dealer (and other comparable CDO desks) are forced to continue buying more and more protection, thereby chasing the market and exacerbating the moves wider.

In practice, the trading dynamics are more complicated but fall along these lines. Dealers are usually buyers of mezzanine tranche protection and hedge by selling single-name CDS. As spreads widen, mezz-tranches become more equity like and mezz tranche hedge-ratios vs single-name CDS tend to decrease as spreads widen. This implies the size of dealers' single name hedges relative to mezz tranches might end up being too great as spreads increase and as a result dealers will need to decrease hedges by buying single-name protection. And vice versa when spreads tighten.

Sphere: Related Content
Print this post


Gentlemutt said...

Nice straightforward piece, Mr. Durden.

This is the exact same phenomenon as so-called Portfolio Insurance in the 1980s, which contributed to the panic selling of stock index futures 'into the hole' during the 1987 stock market crash.

'Dynamic hedging' means one is always chasing the market, and resulted in a lot of profits for interest-rate swap traders at the expense of mortgage bond traders too. At the very least the mortgage traders, by nature always long a portfolio with negative convexity, had to keep crossing the bid/offer as market movements caused them to constantly adjust their hedges.

To use the current meme, desks that take the negative-convexity side, be they CDS desks, mortgage traders, or what-have-you, are betting the tails aren't going to be too fat...and the market isn't going to shimmy too much.

Sometimes they are right. Eat like a bird, sh*t like an elephant.

Anonymous said...

also known as carry-whoring...seems like old news now, but reminds me of all the pain that was inflicted by curve flattening/ inversion in '07-'08...dealers naturally ended up with massive steepeners (also negatively convex) as structured desks sold 6-9yr protection for the extra duration and dealers hedged whatever way they could (i.e. selling 5yr CDS). Then, nary a long-dated bid in sight, all the demand in the world to cover jump, curves collapsed and many a dealer hid under their desks...